Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.8/dft/simd/common/t1fuv_10.c @ 82:d0c2a83c1364

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add FFTW 3.3.8 source, and a Linux build
author Chris Cannam
date Tue, 19 Nov 2019 14:52:55 +0000
parents
children
rev   line source
Chris@82 1 /*
Chris@82 2 * Copyright (c) 2003, 2007-14 Matteo Frigo
Chris@82 3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
Chris@82 4 *
Chris@82 5 * This program is free software; you can redistribute it and/or modify
Chris@82 6 * it under the terms of the GNU General Public License as published by
Chris@82 7 * the Free Software Foundation; either version 2 of the License, or
Chris@82 8 * (at your option) any later version.
Chris@82 9 *
Chris@82 10 * This program is distributed in the hope that it will be useful,
Chris@82 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@82 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
Chris@82 13 * GNU General Public License for more details.
Chris@82 14 *
Chris@82 15 * You should have received a copy of the GNU General Public License
Chris@82 16 * along with this program; if not, write to the Free Software
Chris@82 17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Chris@82 18 *
Chris@82 19 */
Chris@82 20
Chris@82 21 /* This file was automatically generated --- DO NOT EDIT */
Chris@82 22 /* Generated on Thu May 24 08:05:26 EDT 2018 */
Chris@82 23
Chris@82 24 #include "dft/codelet-dft.h"
Chris@82 25
Chris@82 26 #if defined(ARCH_PREFERS_FMA) || defined(ISA_EXTENSION_PREFERS_FMA)
Chris@82 27
Chris@82 28 /* Generated by: ../../../genfft/gen_twiddle_c.native -fma -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name t1fuv_10 -include dft/simd/t1fu.h */
Chris@82 29
Chris@82 30 /*
Chris@82 31 * This function contains 51 FP additions, 40 FP multiplications,
Chris@82 32 * (or, 33 additions, 22 multiplications, 18 fused multiply/add),
Chris@82 33 * 32 stack variables, 4 constants, and 20 memory accesses
Chris@82 34 */
Chris@82 35 #include "dft/simd/t1fu.h"
Chris@82 36
Chris@82 37 static void t1fuv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 38 {
Chris@82 39 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 40 DVK(KP618033988, +0.618033988749894848204586834365638117720309180);
Chris@82 41 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 42 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 43 {
Chris@82 44 INT m;
Chris@82 45 R *x;
Chris@82 46 x = ri;
Chris@82 47 for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@82 48 V T4, TA, Tk, Tp, Tq, TE, TF, TG, T9, Te, Tf, TB, TC, TD, T1;
Chris@82 49 V T3, T2;
Chris@82 50 T1 = LD(&(x[0]), ms, &(x[0]));
Chris@82 51 T2 = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 52 T3 = BYTWJ(&(W[TWVL * 8]), T2);
Chris@82 53 T4 = VSUB(T1, T3);
Chris@82 54 TA = VADD(T1, T3);
Chris@82 55 {
Chris@82 56 V Th, To, Tj, Tm;
Chris@82 57 {
Chris@82 58 V Tg, Tn, Ti, Tl;
Chris@82 59 Tg = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 60 Th = BYTWJ(&(W[TWVL * 6]), Tg);
Chris@82 61 Tn = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 62 To = BYTWJ(&(W[0]), Tn);
Chris@82 63 Ti = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 64 Tj = BYTWJ(&(W[TWVL * 16]), Ti);
Chris@82 65 Tl = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 66 Tm = BYTWJ(&(W[TWVL * 10]), Tl);
Chris@82 67 }
Chris@82 68 Tk = VSUB(Th, Tj);
Chris@82 69 Tp = VSUB(Tm, To);
Chris@82 70 Tq = VADD(Tk, Tp);
Chris@82 71 TE = VADD(Th, Tj);
Chris@82 72 TF = VADD(Tm, To);
Chris@82 73 TG = VADD(TE, TF);
Chris@82 74 }
Chris@82 75 {
Chris@82 76 V T6, Td, T8, Tb;
Chris@82 77 {
Chris@82 78 V T5, Tc, T7, Ta;
Chris@82 79 T5 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 80 T6 = BYTWJ(&(W[TWVL * 2]), T5);
Chris@82 81 Tc = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 82 Td = BYTWJ(&(W[TWVL * 4]), Tc);
Chris@82 83 T7 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 84 T8 = BYTWJ(&(W[TWVL * 12]), T7);
Chris@82 85 Ta = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 86 Tb = BYTWJ(&(W[TWVL * 14]), Ta);
Chris@82 87 }
Chris@82 88 T9 = VSUB(T6, T8);
Chris@82 89 Te = VSUB(Tb, Td);
Chris@82 90 Tf = VADD(T9, Te);
Chris@82 91 TB = VADD(T6, T8);
Chris@82 92 TC = VADD(Tb, Td);
Chris@82 93 TD = VADD(TB, TC);
Chris@82 94 }
Chris@82 95 {
Chris@82 96 V Tt, Tr, Ts, Tx, Tz, Tv, Tw, Ty, Tu;
Chris@82 97 Tt = VSUB(Tf, Tq);
Chris@82 98 Tr = VADD(Tf, Tq);
Chris@82 99 Ts = VFNMS(LDK(KP250000000), Tr, T4);
Chris@82 100 Tv = VSUB(T9, Te);
Chris@82 101 Tw = VSUB(Tk, Tp);
Chris@82 102 Tx = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), Tw, Tv));
Chris@82 103 Tz = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), Tv, Tw));
Chris@82 104 ST(&(x[WS(rs, 5)]), VADD(T4, Tr), ms, &(x[WS(rs, 1)]));
Chris@82 105 Ty = VFNMS(LDK(KP559016994), Tt, Ts);
Chris@82 106 ST(&(x[WS(rs, 3)]), VFNMSI(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@82 107 ST(&(x[WS(rs, 7)]), VFMAI(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@82 108 Tu = VFMA(LDK(KP559016994), Tt, Ts);
Chris@82 109 ST(&(x[WS(rs, 1)]), VFNMSI(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@82 110 ST(&(x[WS(rs, 9)]), VFMAI(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@82 111 }
Chris@82 112 {
Chris@82 113 V TJ, TH, TI, TN, TP, TL, TM, TO, TK;
Chris@82 114 TJ = VSUB(TD, TG);
Chris@82 115 TH = VADD(TD, TG);
Chris@82 116 TI = VFNMS(LDK(KP250000000), TH, TA);
Chris@82 117 TL = VSUB(TE, TF);
Chris@82 118 TM = VSUB(TB, TC);
Chris@82 119 TN = VMUL(LDK(KP951056516), VFNMS(LDK(KP618033988), TM, TL));
Chris@82 120 TP = VMUL(LDK(KP951056516), VFMA(LDK(KP618033988), TL, TM));
Chris@82 121 ST(&(x[0]), VADD(TA, TH), ms, &(x[0]));
Chris@82 122 TO = VFMA(LDK(KP559016994), TJ, TI);
Chris@82 123 ST(&(x[WS(rs, 4)]), VFMAI(TP, TO), ms, &(x[0]));
Chris@82 124 ST(&(x[WS(rs, 6)]), VFNMSI(TP, TO), ms, &(x[0]));
Chris@82 125 TK = VFNMS(LDK(KP559016994), TJ, TI);
Chris@82 126 ST(&(x[WS(rs, 2)]), VFMAI(TN, TK), ms, &(x[0]));
Chris@82 127 ST(&(x[WS(rs, 8)]), VFNMSI(TN, TK), ms, &(x[0]));
Chris@82 128 }
Chris@82 129 }
Chris@82 130 }
Chris@82 131 VLEAVE();
Chris@82 132 }
Chris@82 133
Chris@82 134 static const tw_instr twinstr[] = {
Chris@82 135 VTW(0, 1),
Chris@82 136 VTW(0, 2),
Chris@82 137 VTW(0, 3),
Chris@82 138 VTW(0, 4),
Chris@82 139 VTW(0, 5),
Chris@82 140 VTW(0, 6),
Chris@82 141 VTW(0, 7),
Chris@82 142 VTW(0, 8),
Chris@82 143 VTW(0, 9),
Chris@82 144 {TW_NEXT, VL, 0}
Chris@82 145 };
Chris@82 146
Chris@82 147 static const ct_desc desc = { 10, XSIMD_STRING("t1fuv_10"), twinstr, &GENUS, {33, 22, 18, 0}, 0, 0, 0 };
Chris@82 148
Chris@82 149 void XSIMD(codelet_t1fuv_10) (planner *p) {
Chris@82 150 X(kdft_dit_register) (p, t1fuv_10, &desc);
Chris@82 151 }
Chris@82 152 #else
Chris@82 153
Chris@82 154 /* Generated by: ../../../genfft/gen_twiddle_c.native -simd -compact -variables 4 -pipeline-latency 8 -n 10 -name t1fuv_10 -include dft/simd/t1fu.h */
Chris@82 155
Chris@82 156 /*
Chris@82 157 * This function contains 51 FP additions, 30 FP multiplications,
Chris@82 158 * (or, 45 additions, 24 multiplications, 6 fused multiply/add),
Chris@82 159 * 32 stack variables, 4 constants, and 20 memory accesses
Chris@82 160 */
Chris@82 161 #include "dft/simd/t1fu.h"
Chris@82 162
Chris@82 163 static void t1fuv_10(R *ri, R *ii, const R *W, stride rs, INT mb, INT me, INT ms)
Chris@82 164 {
Chris@82 165 DVK(KP587785252, +0.587785252292473129168705954639072768597652438);
Chris@82 166 DVK(KP951056516, +0.951056516295153572116439333379382143405698634);
Chris@82 167 DVK(KP250000000, +0.250000000000000000000000000000000000000000000);
Chris@82 168 DVK(KP559016994, +0.559016994374947424102293417182819058860154590);
Chris@82 169 {
Chris@82 170 INT m;
Chris@82 171 R *x;
Chris@82 172 x = ri;
Chris@82 173 for (m = mb, W = W + (mb * ((TWVL / VL) * 18)); m < me; m = m + VL, x = x + (VL * ms), W = W + (TWVL * 18), MAKE_VOLATILE_STRIDE(10, rs)) {
Chris@82 174 V Tr, TH, Tg, Tl, Tm, TA, TB, TJ, T5, Ta, Tb, TD, TE, TI, To;
Chris@82 175 V Tq, Tp;
Chris@82 176 To = LD(&(x[0]), ms, &(x[0]));
Chris@82 177 Tp = LD(&(x[WS(rs, 5)]), ms, &(x[WS(rs, 1)]));
Chris@82 178 Tq = BYTWJ(&(W[TWVL * 8]), Tp);
Chris@82 179 Tr = VSUB(To, Tq);
Chris@82 180 TH = VADD(To, Tq);
Chris@82 181 {
Chris@82 182 V Td, Tk, Tf, Ti;
Chris@82 183 {
Chris@82 184 V Tc, Tj, Te, Th;
Chris@82 185 Tc = LD(&(x[WS(rs, 4)]), ms, &(x[0]));
Chris@82 186 Td = BYTWJ(&(W[TWVL * 6]), Tc);
Chris@82 187 Tj = LD(&(x[WS(rs, 1)]), ms, &(x[WS(rs, 1)]));
Chris@82 188 Tk = BYTWJ(&(W[0]), Tj);
Chris@82 189 Te = LD(&(x[WS(rs, 9)]), ms, &(x[WS(rs, 1)]));
Chris@82 190 Tf = BYTWJ(&(W[TWVL * 16]), Te);
Chris@82 191 Th = LD(&(x[WS(rs, 6)]), ms, &(x[0]));
Chris@82 192 Ti = BYTWJ(&(W[TWVL * 10]), Th);
Chris@82 193 }
Chris@82 194 Tg = VSUB(Td, Tf);
Chris@82 195 Tl = VSUB(Ti, Tk);
Chris@82 196 Tm = VADD(Tg, Tl);
Chris@82 197 TA = VADD(Td, Tf);
Chris@82 198 TB = VADD(Ti, Tk);
Chris@82 199 TJ = VADD(TA, TB);
Chris@82 200 }
Chris@82 201 {
Chris@82 202 V T2, T9, T4, T7;
Chris@82 203 {
Chris@82 204 V T1, T8, T3, T6;
Chris@82 205 T1 = LD(&(x[WS(rs, 2)]), ms, &(x[0]));
Chris@82 206 T2 = BYTWJ(&(W[TWVL * 2]), T1);
Chris@82 207 T8 = LD(&(x[WS(rs, 3)]), ms, &(x[WS(rs, 1)]));
Chris@82 208 T9 = BYTWJ(&(W[TWVL * 4]), T8);
Chris@82 209 T3 = LD(&(x[WS(rs, 7)]), ms, &(x[WS(rs, 1)]));
Chris@82 210 T4 = BYTWJ(&(W[TWVL * 12]), T3);
Chris@82 211 T6 = LD(&(x[WS(rs, 8)]), ms, &(x[0]));
Chris@82 212 T7 = BYTWJ(&(W[TWVL * 14]), T6);
Chris@82 213 }
Chris@82 214 T5 = VSUB(T2, T4);
Chris@82 215 Ta = VSUB(T7, T9);
Chris@82 216 Tb = VADD(T5, Ta);
Chris@82 217 TD = VADD(T2, T4);
Chris@82 218 TE = VADD(T7, T9);
Chris@82 219 TI = VADD(TD, TE);
Chris@82 220 }
Chris@82 221 {
Chris@82 222 V Tn, Ts, Tt, Tx, Tz, Tv, Tw, Ty, Tu;
Chris@82 223 Tn = VMUL(LDK(KP559016994), VSUB(Tb, Tm));
Chris@82 224 Ts = VADD(Tb, Tm);
Chris@82 225 Tt = VFNMS(LDK(KP250000000), Ts, Tr);
Chris@82 226 Tv = VSUB(T5, Ta);
Chris@82 227 Tw = VSUB(Tg, Tl);
Chris@82 228 Tx = VBYI(VFMA(LDK(KP951056516), Tv, VMUL(LDK(KP587785252), Tw)));
Chris@82 229 Tz = VBYI(VFNMS(LDK(KP587785252), Tv, VMUL(LDK(KP951056516), Tw)));
Chris@82 230 ST(&(x[WS(rs, 5)]), VADD(Tr, Ts), ms, &(x[WS(rs, 1)]));
Chris@82 231 Ty = VSUB(Tt, Tn);
Chris@82 232 ST(&(x[WS(rs, 3)]), VSUB(Ty, Tz), ms, &(x[WS(rs, 1)]));
Chris@82 233 ST(&(x[WS(rs, 7)]), VADD(Tz, Ty), ms, &(x[WS(rs, 1)]));
Chris@82 234 Tu = VADD(Tn, Tt);
Chris@82 235 ST(&(x[WS(rs, 1)]), VSUB(Tu, Tx), ms, &(x[WS(rs, 1)]));
Chris@82 236 ST(&(x[WS(rs, 9)]), VADD(Tx, Tu), ms, &(x[WS(rs, 1)]));
Chris@82 237 }
Chris@82 238 {
Chris@82 239 V TM, TK, TL, TG, TO, TC, TF, TP, TN;
Chris@82 240 TM = VMUL(LDK(KP559016994), VSUB(TI, TJ));
Chris@82 241 TK = VADD(TI, TJ);
Chris@82 242 TL = VFNMS(LDK(KP250000000), TK, TH);
Chris@82 243 TC = VSUB(TA, TB);
Chris@82 244 TF = VSUB(TD, TE);
Chris@82 245 TG = VBYI(VFNMS(LDK(KP587785252), TF, VMUL(LDK(KP951056516), TC)));
Chris@82 246 TO = VBYI(VFMA(LDK(KP951056516), TF, VMUL(LDK(KP587785252), TC)));
Chris@82 247 ST(&(x[0]), VADD(TH, TK), ms, &(x[0]));
Chris@82 248 TP = VADD(TM, TL);
Chris@82 249 ST(&(x[WS(rs, 4)]), VADD(TO, TP), ms, &(x[0]));
Chris@82 250 ST(&(x[WS(rs, 6)]), VSUB(TP, TO), ms, &(x[0]));
Chris@82 251 TN = VSUB(TL, TM);
Chris@82 252 ST(&(x[WS(rs, 2)]), VADD(TG, TN), ms, &(x[0]));
Chris@82 253 ST(&(x[WS(rs, 8)]), VSUB(TN, TG), ms, &(x[0]));
Chris@82 254 }
Chris@82 255 }
Chris@82 256 }
Chris@82 257 VLEAVE();
Chris@82 258 }
Chris@82 259
Chris@82 260 static const tw_instr twinstr[] = {
Chris@82 261 VTW(0, 1),
Chris@82 262 VTW(0, 2),
Chris@82 263 VTW(0, 3),
Chris@82 264 VTW(0, 4),
Chris@82 265 VTW(0, 5),
Chris@82 266 VTW(0, 6),
Chris@82 267 VTW(0, 7),
Chris@82 268 VTW(0, 8),
Chris@82 269 VTW(0, 9),
Chris@82 270 {TW_NEXT, VL, 0}
Chris@82 271 };
Chris@82 272
Chris@82 273 static const ct_desc desc = { 10, XSIMD_STRING("t1fuv_10"), twinstr, &GENUS, {45, 24, 6, 0}, 0, 0, 0 };
Chris@82 274
Chris@82 275 void XSIMD(codelet_t1fuv_10) (planner *p) {
Chris@82 276 X(kdft_dit_register) (p, t1fuv_10, &desc);
Chris@82 277 }
Chris@82 278 #endif